BROOKE GLADSTONE: This is On the Media. I'm Brooke Gladstone.

BOB GARFIELD: And I'm Bob Garfield. This is poll season, and there's no vaccine. We've all been infected -- not that there's anything wrong with that.

FRANK NEWPORT: You know, in a democracy, the people are the source of wisdom and ultimate power, and I would like my elected representative to poll all the time, trying to understand that wisdom of the people.

BOB GARFIELD: Yeah, well, that's Frank Newport, editor in chief of the Gallup Poll, so he would say that. But there's no separating polling from politicking these days, so despite its hallowed place in the pantheon of pollsters, this year Gallup is being charged with a Republican bias, because it assumes that only 15 percent of voters will be non-white, and 18 percent will be liberal. (Both were a few percentage points higher in 2000, and we just don't know what the political and racial complexion of the electorate will be this time around.)

BROOKE GLADSTONE: The conflicting methodologies of various pollsters, from ABC to Zogby, have led to election forecasts that can be pretty much anything you want them to be. And the media, which carry the poll parasite like a swarm of tse-tse flies, are the first to admit it. [TAPE PLAYS]

WOLF BLITZER: Four years ago, roughly two weeks before election day, a CNN/USA Today Gallup tracking poll of likely voters would continue to fluctuate, though usually showing Bush ahead, sometimes by as much as 13 points. But it was Gore who went on to win the popular vote. [TAPE ENDS]

BROOKE GLADSTONE: But for those grown skeptical of those countless, conflicting, unreliable polls, CNN has engineered something altogether new for voters to contend with. [TAPE PLAYS]

WOLF BLITZER: But our poll of polls, which averages major surveys including our own, shows a tight race. [TAPE ENDS]

BROOKE GLADSTONE: That's right. The poll of polls. CNN is hardly alone, though. The race to predict is breeding more polls, not less. Philip J. Trounstine, director of the Survey and Policy Research Institute at San Jose State University, wrote a primer on polling for the Los Angeles Times this week, and he joins me now. Welcome to the show.

PHILIP J. TROUNSTINE: It's a pleasure to be here.

BROOKE GLADSTONE: So how can so many news-gathering organizations with such faith in their own polling methods come up with such different results.

PHILIP J. TROUNSTINE: There are different methodologies involved. How they choose likely voters, question order, how questions are worded and what order they take place in, and of course when polling is done.

BROOKE GLADSTONE: Take, for instance, two very different methodologies -- that used by Zogby and that used by the Gallup polls -- tell me how they could lead to different results.

PHILIP J. TROUNSTINE: Zogby asks people what their party identification is, and if he comes out with 50 percent Democrats and 30 percent Republican, he'll balance that survey -- that is, by applying a weight, so that he has the same proportion of Republicans and Democrats that there were in the 2000 election, as measured by exit polls. Other pollsters, like Gallup, believe that party identification - that it's an attitude -they count on the randomness of their survey to give them a representative sample, and they don't try to balance. So they might wind up with 40 percent Republicans and 35 percent Democrats.

BROOKE GLADSTONE: What do you think of the recent criticism of Gallup that it tends to under-estimate the black electorate and also under-estimate the liberal electorate, and therefore their polls are coming out far more strongly in favor of Bush than is actually the case?

PHILIP J. TROUNSTINE: Well, I think that's a criticism that's primarily leveled by people who are supporters of Kerry. The Gallup poll has a magnificent record. I think there's a question that can be raised about the methodology. They have a 7-point scale that they use, and people are considered likely voters if they score high enough on that scale. That includes questions like did they vote in the last election, how closely are they following the race, do they know where their polling place is. Some people argue that this under-states lower-income blacks, younger people. Gallup would certainly argue that based on the historical record, their methodology has been extremely reliable.

BROOKE GLADSTONE: Let's talk about the new and probably unresolvable crisis in polling which is the cell phone. More and more Americans are using cell phones as their only phone, especially young people. And there really is currently no adequate way to poll this unrepresented group.

PHILIP J. TROUNSTINE: If you take a look at some of the states that are really in play this year, these are states with big university populations, and let me give you an example of what that might mean. I lectured in a class at San Jose State last week. Virtually everyone in the class is a registered voter because the professor's pushing them to be registered voters. Twenty-five percent of the people, when I asked for a show of hands, did not have a land line. They have only a cell phone. Those people are not in any poll. At the same time that group of people tends not to be voters in very big numbers.

BROOKE GLADSTONE: Let's talk briefly about the margin of error. I see, again and again, on CNN, in newspapers, in blogs -- they often suggest that if the candidates are more than 4 points apart, then the one who's ahead is actually beyond the margin of error, and we know that's not true. You really have to have an 8 point difference to overcome a 4 point margin of error.

PHILIP J. TROUNSTINE: You see a lot of stories where reporters say Bush is ahead 49 to 48. Those numbers don't mean much, except in the aggregate, except when you look at lots and lots of polls, which show a 3 and 4 point lead. Then you can pretty well guess that Bush has a slight lead.

BROOKE GLADSTONE: If you have a public consuming these polls all the time, the public can get drawn up into the numbers and, as a result, they can become, in some cases, a self-fulfilling prophecy?

PHILIP J. TROUNSTINE: It's possible, but I don't think that's a terrible problem. We can assume that people are bright enough to be able to figure out that these are snapshots in time and they're not necessarily guides to how they should act in the future.

BROOKE GLADSTONE: But I guess you can forgive those of us in the public for hoping for a little schadenfreude when the headline "Dewey Wins" comes out based on the polls, and it isn't Dewey after all.

PHILIP J. TROUNSTINE: The cell phone issue and how pollsters have chosen their likely voters are the two pieces of the puzzle that make possible a "Dewey Beats Truman" event. And I think there'll be plenty of people out there who would say --"Ah-ha! Boy, am I glad those pollsters messed it up."

BROOKE GLADSTONE: [LAUGHS] Thank you very much.

PHILIP J. TROUNSTINE: You're welcome.

BROOKE GLADSTONE: Philip J. Trounstine is director of the Survey and Policy Research Institute at San Jose State University.